Causal analysis of Canada’s environment-growth nexus for inclusive development metrics

Little is known about the relevance of alternative measures of growth in environmental and developmental economic analyses. In Canada, for example, no literature exists on whether there is a causal link between the level of environmental performance and alternative measures of economic progress (which are argued to better capture the overall economic wellbeing than the gross domestic product—GDP). As environmental policies may relate to overall economic wellbeing, we opine that understanding overall economic progress is essential for achieving sustainable development and emissions reduction targets. This paper addresses a knowledge gap by assessing the causal links and directions between Canada’s national-level greenhouse gas emissions (GHG—as an indicator of environmental performance) and three alternative measures of economic growth, namely, gross national disposable income (GNDI), human development index (HDI), and index of economic freedom (IEF); from 1995 to 2019. Our results indicate that causality exists between Canada’s GHG and the alternative growth measures. This implies that Canada’s GNDI, HDI, and IEF may be useful and complementary to GDP in forecasting the national-level total GHG emissions. The research provides insights to further consider the role of overall economic wellbeing in the quest for sustainable, lower-emissions, economic development in Canada and by extension in other nations.


Introduction
Although the literature on alternative measures of growth (also referred to as inclusive development metrics, beyond GDP-BGDP, or GDP-alternative measures) is limited, recent studies are beginning to demonstrate the potentially pivotal role of economic wellbeing on understanding and predicting aspects of the overall societal progress (Kimmerer 2020, Miller et al 2021b, Kamran et al 2023).In Canada, for example, it has been documented that a long run relationship exists between Canada's nationallevel per capita green house gas emission (GHGpcconsidered an indicator of environmental quality) and three inclusive development measures, namely gross national disposable income (GNDI), human development index (HDI), and the index of economic freedom (IEF) (Iwuoha andOnochie 2022, 2023).
We have observed, however, that no research exists that determines whether there is causality between Canada's GHGpc and inclusive growth indicators, and the causal directions in Canada's environment-growth nexus using GHGpc and BGDP measures.As a result, it is unclear if GDP-alternative measures can be used to forecast Canada's GHGpc.We believe that policy makers would be better informed if causality and causal directions between Canada's GHGpc and growth measures are determined, thus creating access to a more holistic body of knowledge on the potential role of societal wellbeing in the quest to achieving sustainability targets such as green house gas (GHG) emissions reduction.
1.1.Summary of the literature Causal analysis techniques are common practice in energy, environment, and development economics literature for evaluating short and longrun causality between variables (Baranzini et al 2013, Omri et al 2014, Wang et al 2016, Mirza and Kanwal 2017, Teng et al 2020).Yildirim and Aslan (2012) studied the environment-growth nexus of seventeen Organization for Economic Cooperation and Development (OECD) countries by investigating energy consumption, employment, gross fixed capital formation, and GDP.Their bootstrapcorrected causality tests showed mixed causality results with unidirectional causality running from GDP to energy for Canada, Australia, and Ireland.Although bi-directional causality was observed running from energy consumption to the real GDP for Spain, Norway, Italy, and New Zealand, the tests only revealed unidirectional causality for Japan.The authors, therefore, proposed that for certain OECD countries (e.g.Japan, Italy, Norway, and New Zealand) an aggregated level approach to environmental policy should not be adopted.
Thirty OECD countries were included in the comprehensive analysis of the energy consumption and growth of 119 countries performed by Ahmed and Azam (2016).The authors applied the Grangercausality analysis in a frequency domain context to better understand the short-run and long-run causality between the variables.Their study indicated that birectionality, unidirectional Granger, and neutrality exists for different high-income OECD countries between energy consumption and economic growth.As such, the authors opined that targeted and strategic policy development is required to capture both short and long-run causal dynamics in the energy and sustainable development nexus.
Destek (2016) applied panel dynamic ordinary least squares (DOLS) and fully modified ordinary least squares (FMOLS) regression in their data analysis of twenty-six OECD countries between 1991 and 2013 and determined that natural gas consumption in OECD countries is positively affected by GDP growth in the long run.The study utilized vector error correction model (VECM) Granger causality test to reveal that the causal direction from natural gas consumption to GDP growth is unidirectional in the short-run and bi-directional in the long run.
Hamit-Haggar (2012) used data from 1991 to 2007 to analyze the long-run and causal relationship between GHG emissions, energy consumption, and growth for several Canadian industries.Granger causality was run on disaggregate panel level data while accounting for cross-sectional dependencies.A weak one-way causality flowing from economic growth to GHG emissions was confirmed in the long run.In the short run, unidirectional causality ran from economic growth to GHG emissions.
Ghali and El-Sakka (2004) performed Vector Error Correction Model (VECM) Granger causality tests on Canada's energy use and output growth.They found that bi-directional Granger ran from energy use to economic growth.This study considered output, labour, capital, and energy use as separate inputs but did not capture any BGDP growth indicators.
Despite the preponderance of literature on causality testing in the energy, environment-growth body of knowledge, assessments of causality and causal directions using GDP-alternative measures are limited and less-readily accessible.As a result, in the case of Canada where even more limited-to-no BGDP-based causal analysis exists, it is unclear if there are causal links between the level of environmental performance and alternative measures of economic progress, and the direction of causality.

Data and methodology
This paper uses GHG data for the 1995-2019 period from the United Nations (UN), along with GDP and inclusive development measures (GNDI, HDI, and IEF) from the OECD, United Nations Development Programme (UNDP), and the Heritage Foundation databases (Government of Canada 2020, OECD 2020, UNDP 2020, Miller et al 2021a, Iwuoha and Onochie 2022, 2023).
Data on inclusive development metrics remains limited both for Canada and other countries (Kimmerer 2020, Miranda et al 2020).The three BGDP datasets selected for this study were based on their availability for Canada as well as the reliability of the data source.Kimmerer (2020) catalogued the application of 40+ BGDP measures for environmental, developmental, and sustainability studies, arguing for the use of inclusive growth measures beyond income and output-based metrics.Miranda et al (2020) studied Canada's HDI with carbon dioxide emissions from 1990-2016.No studies exist to date (to the best of our knowledge) for the three BGDP datasets (GNDI, HDI, and IEF) with total GHG emissions.
Vector autoregression (VAR) is described as a stochastic model used to capture the linear interdependencies among multiple time series (Bella et al 2014, Sebri andBen-Salha 2014).Previous Johanssen cointegration tests of Canada's national-level GHG with the inclusive development measures GNDI, HDI, and IEF indicate that a long-run relationship exists between GHG and the growth measures, with GHG not being weakly exogenous in the cointegrating relationship (Iwuoha and Onochie 2022).This means that the variables being evaluated individually have non-stationary unit roots (i.e., integrated of the order 1 or I(1)), with a unit root-stationary (i.e.I(0)) linear combination that exists between the variables.We note that where the variables are integrated of the order 0 (i.e.I(0)) or are stationary, co-movement may not exist between GHG and the inclusive delevelopment measures, thus obviating the need for cointegration tests or linear interdependency studies.The cointegration between Canada's national-level GHG and GNDI, HDI, and IEF permit further linear interdependency-type analyses of the GHG-inclusive growth measures nexus as performed in this paper.
The VAR models generated were set up using the first differences of the variables in both the 'constant ' and'constant andtrend' scenarios (Iwuoha andOnochie 2022, 2023).Both VAR scenarios incorporated heteroskedasticity and autocorrelation consistent (HAC) standard error estimation.Lag order specification was evaluated based on three information criteria, Akaike Information Criterion (AIC), Schwarz's Bayesian Criterion (BIC), and the Hannan-Quinn information Criterion (HQC).The lag order selected was based on the minimized AIC which also had the least p-value from the Likelihood Ratio (LR) test.
Given that VAR models suffer from limitations based on their assumptions of linearity and stationarity, and may not capture true causality, post-model specifications tests were completed and indicated that for the data and time series period evaluated in this paper, the model specification was adequate, heteroskedasticity and autoregressive conditional heteroskedasticity (ARCH) are not present, the model error was normally distributed, there was no autocorrelation nor changes in the parameters.
Causality  2020).Granger causality and causal directions were determined using the p-values of the F-statistics of the VAR models.The null hypothesis (H0) was that Granger causality did not exist between the variables.Failure to reject H0 occurred if the p-value is >0.05 and significant, implying the absence of Granger causality running from the independent variable to the dependent variable for the VAR equation being evaluated (Fallahi 2011, Pao and Tsai 2011, Vaona 2012, Gorus and Aydin 2019).

Results
The summary of the VAR test statistic results and the inferred Granger causality is presented in tables 1 and 2 below.

Discussion and implications
For GDP, the more commonly evaluated growth indicator in the literature (Destek et al 2020, Okumus and Bozkurt 2020), the paired-variables VAR model result that corresponds to the 'constant' scenario suggests that bi-directional Granger causality exists between GHGpc and GDPpc (tables 1 and 2), indicating the existence of a feedback effect between both variables.In the 'constant and trend' scanario, however, GDP Granger causes GHGpc, suggesting that increasing output trends could directly impact environmental performance.
Moving on to the inclusive growth measures, GNDIpc-GHGpc causality results suggest a feedback effect in the 'constant' scenario, whereas in the 'constant and trend' scenario, a unidirectional Granger runs from GHGpc to GNDIpc.This may mean that increases in the national-level total emissions could lead to economic side effects that can impact the levels of disposable income.The finding demonstrates that environmental management and emissions-reductions initiatives can directly affect the overall economic wellbeing and should, therefore, be earnestly pursued as being beneficial for economic progress.
HDI Granger causes GHGpc in both scenarios tested.The unidirectional Granger that runs from HDI to GHGpc suggests that improving human development could lead to increased emissions.This underscores that the level of human development should be given thoughtful consideration as emissions reduction policies are developed and implemented in Canada.The observation further buttresses the need for policies and strategies for managing and achieving environmental targets to also account for other aspects of the society other than income and output.
IEF has a feedback effect with GHGpc in both scenarios tested.This can be considered indicative of the importance of economic freedom in the society as efforts are pursued to reduce Canada's GHG emissions.Simply put, for policy makers and sustainability planners, the level of economic freedom should not be overlooked during strategy development and implementation.The lesson learned from evaluating IEF is a complementary insight that may not have been easily gained if relying solely on income and output-based growth measures.
Although causality studies using inclusive development metrics are limited, observations from this study that are related to the more generally used income and output growth metrics can be contrasted with the Granger causality literature.For example, Soytas et al (2007) indicated that in the United States of America (USA), income   2. (e) Except where indicated, the VARs were set up using a lag order of 7 obtained from the paired variables VAR lag selection.(f) * * * * * For this scenario, a lag order of 6 was used as model convergence was not obtained using the lag of 7 from the VAR lag selection (also see table 2).(g) First difference variables are reported using a 'd' prefix.
(h) pc refers to per capita values of the variable.1. (e) * * * * * A lag order of 6 was used as model convergence was not obtained using the lag of 7 from the VAR lag selection.
does not Granger cause environmental degradation (proxied by CO 2 emisisons).Our study, however, by evaluating GNDIpc (albeit based on Canada) provides a contrasting view that highlights the potential for disposable income (an inclusive growth metric) to affect the national-level total emissions.Dogan and Turkekul (2016) found in their USA country study that casaulity exists between GDP and CO 2 emissions.Our paper, though based on Canada, supports their observation that both environmental performance and growth could affect each other.
For BGDP measures evaluation in Canada, the findings from both scenarios tested in this paper suggesting that increasing HDI could lead to increase in total GHG emissions agrees with Miranda et al (2020), who found that improvement in HDI trends would lead to decreased national-level environmental performace (measured, howbeit, by CO 2 emissions) (Miranda et al 2020).
The findings from evaluating both the 'constant' and 'constant and trend' scenarios provided insight on the potential or extent to which the reciprocal relationship between growth and environmental quality can affect the VAR estimates and consequently the causality and causal direction interpretations, thereby providing a 'broad' context of possible results for policy design considerations.
It is worth mentioning that, as emissions and growth increasingly decouple (due to factors such as the proliferation of lower cost, lower carbon energy sources, structural economic differences or evolutions such as the shift from a manufacturing to a service-based economy, and increasing energy efficiency and electrification, population/demographic changes among others) going forward, the level of relevance of Granger causality studies of GHGpc and the evaluated BGDP measures may reduce over time.This may necessitate the use of non growth-related explanatory measures.

Conclusion
The use of causality results can inform environmental as well as social and sustainability policy development, especially if the direction of causality is determined.At a macro level in Canada, by highlighting the existence of causal links between the BGDP metrics and the environment (GHG), this research supports the theory that societal interactions and progress are dynamically linked with (i.e. they do not occur in isolation from) the environment (Seto et al 2016).The degree of interaction of the wider-scale economic well-being with environmental quality may vary from one jurisdiction to another.Therefore, measuring, understanding, and managing the wider-scale economic progress would be essential for ultimately achieving climate (i.e.GHG emission reduction) targets.
In the light of the above, the implication of this study is to inform policy design, implementation, and monitoring.At the end of the day, sustainability should be comprised of the follow-through over time of policy performance to ensure that policies remain agile and continue to address the needs of our dynamic societies and economies.
Given the focus of this study on the national-level GHG emissions, we underscore that care should be taken when considering the applicability of the results of this assessment in Canada at regional, provincvial, and territorial levels due to sector-specific or regional emission variations.Different sectors, such as energy, transportation, and agriculture, have diverse impacts on GHG emissions, necessitating targeted policy recommendations for effective reduction.Canada's diverse climate and industries result in significant regional emission disparities which could potentially lead to less effective policy recommendations if designed using results from a national-level study.
Should an interest arise to apply the detailed results from this research to another country, it is important to consider that the limited use of inclusive development metrics in the literature and the focus of this research on Canada's national-level total GHG emissions, may limit the precise application of the results.Other omitted variables (which would vary from one country to another) such as energy (domestic resources, prices, intensity, and consumption), trade dynamics, and environmental policy instruments (e.g.carbon pricing/cap and trade), amongst others, may share links with BGDP measures and as such, play a contributory role in societal wellbeing that could ultimately affect environmental quality.We believe these omitted variables are worth evaluating in subsequent studies both in Canada and in other jurisdictions to gain a more holistic view of GHG causality and the causal directions.
With the above said, what may, however, be generalizable by policy makers and sustainability planners, could be the analytical approach adopted and variables evaluated in this research which extend the existing body of knowledge.This study is all the more pivotal, given the inadequate understanding of the GHG-BGDP nexus and the limited knowledge of the causal relationships between GHG and inclusive measures of economic development.
evidence from a panel of BRIC (Brazil, Russian Federation, India, and China) (b) * * This is the p-value of the F-statistic of the independent variable in the VAR equation being evaluated.(c) * * * Denotes a unidirectional Granger causality running from GDPpc to GHGpc.(d) * * * * Denotes a unidirectional Granger causality running from GHGpc to GDPpc.(e) The unidirectional arrow representations in (c) and (d) are retained for the forward and backward Granger causality directions for all the Granger causality directions reported in this paper.The arrow indicator for bi-directional Granger is provided in table tests are carried out to determine if a time series can be used to forecast another series (Kumar et al 2012).Granger causality is a statistical hypothesis test that leverages VAR model results and is well-documented in the literature (Granger 1969, Destek 2016, Esso and Keho 2016, Bekhet et al 2017, Chaabouni and Saidi 2017, Dogan and Aslan 2017, Ouyang and Li 2018, Bekun et al 2019, Teng et al

Table 1 .
VAR models with test statistic and Granger causality for paired GHGpc and growth indices.
Note: (a) * The equation arrangement to report the results in this table is dependent/independent variable.The VARs were set up using the variable first differences which are stationary.

Table 2 .
Summary of the Granger causal directions for paired GHGpc and growth indices.Refers to the VAR relationship between both variables.The reported result summarizes the Granger causality and causal directions observed from both paired variable equations in table 1. (c) * * Denotes a bi-directional Granger causality running from GDPpc to GHGpc and from GHGpc to GDPpc (feedback effect).The bi-directional arrow in this table was retained to represent the feedback effect Granger causality direction for all the reported instances of bi-directional Granger causality.(d) The arrow indicator for unidirectional Granger causality is provided in table Note: (a) This table is summarized from table 1.(b) * countries Energy 36 685-93 Sebri M and Ben-Salha O 2014 On the causal dynamics between economic growth, renewable energy consumption, CO2 emissions and trade openness: fresh evidence from BRICS countries Renew.Sustain.Energy Rev. 39 14-23 Seto K C, Davis S J, Mitchell R B, Stokes E C, Unruh G and Ürge-Vorsatz D 2016 Carbon lock-in: types, causes, and policy implications Annu.Rev. Environ.Resour.41 425-52 Soytas U, Sari R and Ewing B T 2007 Energy consumption, income, and carbon emissions in the United States Ecol.Econ.62 482-9 Teng T, Ortiz J, Chuanhua X and Fangjhy L 2020 Economic growth, energy consumption, and carbon dioxide emissions in the E7 countries: a bootstrap ARDL bound test Energy Sustain.Soc. 10 20 UNDP 2020 UNDP Human Development Reports 2020: human Development Index (HDI) (United Nations Development Programme) (available at: http://hdr.undp.org/en/indicators/137506) Vaona A 2012 Granger non-causality tests between (non)renewable energy consumption and output in Italy since 1861: the (ir)relevance of structural breaks Energy Policy 45 226-36 Wang K, Zhu B, Wang P and Wei Y-M 2016 Examining the links among economic growth, energy consumption, and CO2 emission with linear and nonlinear causality tests Nat.Hazards 81 1147-59 Yildirim E and Aslan A 2012 Energy consumption and economic growth nexus for 17 highly developed OECD countries: further evidence based on bootstrap-corrected causality tests Energy Policy 51 985-93